Aimed at the problems in the SNR estimate method of the subspace that error is excessive in calculating signal subspace dimension and this makes the estimate precision low under conditions of low SNR, in this paper, a new robust SNR estimate method is proposed applicable to the subspace SNR estimate of small size data through the analysis, the simulation, and the determination of the subspace method under conditions of the finite sample date. The results show that the estimate error can be decreased by 0.3~2 dB by use of the new method in low SNR and short data samples environment